[
https://issues.apache.org/jira/browse/SPARK-30474?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zaisheng Dai updated SPARK-30474:
---------------------------------
Summary: Writing data to parquet with dynamic partitionOverwriteMode should
not do the folder rename in commitjob stage (was: Writing data to parquet with
dynamic partition should not be done in commit job stage)
> Writing data to parquet with dynamic partitionOverwriteMode should not do the
> folder rename in commitjob stage
> --------------------------------------------------------------------------------------------------------------
>
> Key: SPARK-30474
> URL: https://issues.apache.org/jira/browse/SPARK-30474
> Project: Spark
> Issue Type: Improvement
> Components: Input/Output
> Affects Versions: 2.3.4, 2.4.4
> Reporter: Zaisheng Dai
> Priority: Minor
>
> In the current spark implementation if you set,
> {code:java}
> spark.sql.sources.partitionOverwriteMode=dynamic
> {code}
> even with
> {code:java}
> mapreduce.fileoutputcommitter.algorithm.version=2
> {code}
> it would still rename the partition folder *sequentially* in commitJob stage
> as shown here:
> [https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/internal/io/HadoopMapReduceCommitProtocol.scala#L188]
>
> [https://github.com/apache/spark/blob/branch-2.4/core/src/main/scala/org/apache/spark/internal/io/HadoopMapReduceCommitProtocol.scala#L184]
>
> This is very slow in cloud storage. We should commit the data similar to
> FileOutputCommitter v2?
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
